The idea is simple (and it was suggested to me by Ludovico Failla, brilliant musician and Max MSP hacker): downsample a generated sound and use the data to change its parameters.
You can get interesting and unexpected results (the type I prefer!)
Here is an oscillator which is frequency-shifted using the oscillator output as a parameter. Later I use the frequency shifter itself to change the parameter (i.e. I create a feedback between the sound output and the parameter manipulation).